Second, it doesn’t alleviate enough pressure; the bureaucrats can’t tell real solutions from bad ones; the cost to build an unaligned AGI drops each year; etc., etc.
And making AI researchers answerable to some centralized entity puts a big target on that entity as a thing for sociopaths (broadly construed) to corrupt / capture, and the more that AI researchers are living under the thumb of a corrupt / captured entity, the more they’re misaligned with their own values and less likely to be sane.
And making AI researchers answerable to some centralized entity puts a big target on that entity as a thing for sociopaths (broadly construed) to corrupt / capture, and the more that AI researchers are living under the thumb of a corrupt / captured entity, the more they’re misaligned with their own values and less likely to be sane.